1. Field of the Invention
The present invention relates to a multiple-input multiple-output (MIMO) wireless communication method for communications between a sending or transmission terminal having a plurality of antennas and a receiving or reception terminal having a plurality of antennas and a MIMO wireless communication apparatus for communications by the MIMO wireless communication method. More particularly, the present invention relates to a MIMO wireless communication method and apparatus for communications at a high transmission rate if a signal-to-noise (SNR) ratio is high for an eigenmode transmission method which is one type of the MIMO communication method.
2. Description of the Related Art
A communication method has advanced toward a large capacity because of recent expansion of communication demands. This trend is remarkable also in wireless communications. For example, a communication capacity expands also in wireless LAN standards IEEE802.11 defined by Institute of Electrical and Electronics Engineers, Inc (IEEE) which standardizes Local Area Networks (LAN). A communication capacity was initially set to 2 Mbps in IEEE802.11, then expanded to a maximum of 11 Mbps in IEEE802.11b and a maximum of 54 Mbps in IEEE802.11a and 11g, and is expected to be set to a maximum of 600 Mbps in IEEE802.11n whose standardization will be completed in 2007.
MIMO technologies are adopted in IEEE802.11n and the like as an approach to realizing a large capacity of wireless communications. FIG. 2 is a schematic diagram showing a MIMO wireless communication system. A transmission terminal has N transmission antennas 202 and a reception terminal has N reception antennas 203. A transmission antenna signal vector t is defined by a formula (1) by representing signals transmitted from the transmission antennas 202-1 to 202-N by t1 to tN, respectively:
  t  =      (                                        t            1                                                ⋮                                                  t            N                                )  
Similarly, a reception antenna signal vector r is defined by a formula (2) by representing signals received at the reception antennas 203-1 to 203-N by r1 to rN, respectively:
                    r        =                  (                                                                      r                  1                                                                                    ⋮                                                                                      r                  N                                                              )                                                
Transformation from t to r can be expressed by a linear transformation of a formula (3):r=Ht.  (3)
A matrix H representative of this linear transformation is called a channel matrix. Noises are generated at the same time in an actual case. Therefore, noise components n are added as in a formula (4):r=Ht+n  (4)
The channel matrix can be estimated at the reception terminal by transmitting a known signal from the transmission terminal to the reception terminal. This is called channel matrix estimation, and the transmitted known signal is called a training signal. The channel matrix estimation is conducted before MIMO wireless communications are performed.
In FIG. 2, transmission data signals are represented by x1 to xN, reception data signals are represented by y1 to yN, a transmission data signal vector x is defined by a formula (5), and a reception data vector y is defined by a formula (6):
      x    =          (                                                  x              1                                                            ⋮                                                              x              N                                          )            y    =          (                                                  y              1                                                            ⋮                                                              y              N                                          )      
A transmission antenna weight unit 201 in the transmission terminal transforms x into t by linear transformation. A reception antenna weight unit 204 in the reception terminal transforms r into y by linear transformation. The simplest method of realizing MIMO wireless communications is a zero-forcing (ZF) method of using a transmission antenna weight as a unit matrix and a reception antenna weight as an inverse matrix of H. According to the ZF method, a relation between x and y is expressed by a formula (7):
                    y        =                ⁢                              H                          -              1                                ⁢          r                                        =                ⁢                              H                          -              1                                ⁡                      (                          Ht              +              n                        )                                                  =                ⁢                  t          +                                    H                              -                1                                      ⁢            n                                                  =                ⁢                  x          +                                    H                              -                1                                      ⁢            n                              
By cancelling H by its inverse matrix, it becomes possible to recover the transmission data signal at the reception terminal. However, noises are amplified by the inverse matrix of H.
Apart from the ZF method, there is a method called an eigenmode transmission method of realizing MIMO wireless communications. First, this method performs singular value decomposition (SVD) of H expressed by a formula (8):H=USVH)
S is a diagonal matrix whose all elements are positive real numbers, and U and V are unitary matrices. A superscript H of V means Hermitian conjugate (=transpose+complex conjugate). Diagonal elements of S are called singular values. It is assumed that elements of S are called a first singular value, a second singular value, . . . starting from the upper left element, and arranged in the order of larger singular values. The transmission antenna weight is represented by V and the reception antenna weight is represented by Hermitian conjugate of U. A relation between x and y in the eigenmode transmission method is expressed by a formula (9):
                    y        =                              U            H                    ⁢          r                                        =                              U            H                    ⁡                      (                          Ht              +              n                        )                                                  =                              U            H                    ⁡                      (                                                            USV                  H                                ·                Vx                            +              n                        )                                                  =                  Sx          +                                    U              H                        ⁢            n                              
This utilizes the property that Hermitian conjugate of a unitary matrix is equal to the inverse matrix. A signal multiplying the transmission data signal by the singular value is obtained separately from the reception data signal when considering that the amplitude of noises n at the last term in the formula (9) does not change at all because of the property of the unitary matrix and that S is the diagonal matrix. Data communications from x1 to y1 are called a first eigenmode, and sequentially thereafter data communications are called a second eigenmode, a third eigenmode, . . . . In each eigenmode, a transmission gain (loss if smaller than 1) of a square of the singular value is obtained. It is commonly known that the eigenmode transmission method is the MIMO wireless communication method capable of realizing the largest communication capacity.
However, as different froth the ZF method, the eigenmode transmission method is required to set the transmission antenna weight to the matrix calculated from H. Further, since H is estimated at the reception terminal, it is necessary for the reception terminal to feed back information on H to the transmission terminal. Therefore, information is required to be transferred as shown in FIG. 3. It is to be noted that both the transmission and reception terminals have a transmission and reception function. First, the transmission terminal transmits training data, and the reception terminal receives it and estimates the channel matrix. The channel matrix is returned to the transmission terminal which in turn determines the transmission antenna weight from SVD of the channel matrix. Next, after executing a transmission antenna weight process, the transmission terminal transmits a training signal, and the reception terminal estimates again the channel matrix from the received training signal to determine the reception antenna weight. Thereafter, the transmission terminal transmits a transmission data signal subjected to the transmission antenna weight process, and the reception terminal recovers data by the reception antenna weight to thereby establish data communications. Although the transmission terminal determines the transmission antenna weight by SVD as shown in FIG. 3, the transmission antenna weight may be determined at the reception terminal as shown in FIG. 4. In this case, the transmission antenna weight is fed back.
In FIGS. 3 and 4, first SVD does not determine the reception antenna weight, but the reception antenna weight is separately determined by obtaining the channel matrix by using the training signal subjected to the transmission antenna weight process. A channel matrix H′ to be estimated in this case is expressed by a formula (10):
                              H          ′                =        HV                                =                              USV            H                    ·          V                                        =        US            
If the ZF method is used for determining the reception antenna weight, a reception antenna weight R is expressed by a formula (11):
                    R        =                  H                      ′            -            1                                                  =                              (            US            )                                -            1                                                  =                              S                          -              1                                ⁢                      U            H                              
Therefore, a relation between x and y can be expressed by a formula (12):
                    y        =                              S                          -              1                                ⁢                      U            H                    ⁢          r                                        =                              S                          -              1                                ⁢                                    U              H                        ⁡                          (                              Ht                +                n                            )                                                              =                              S                          -              1                                ⁢                                    U              H                        ⁡                          (                                                                    USV                    H                                    ·                  Vx                                +                n                            )                                                              =                  x          +                                    S                              -                1                                      ⁢                          U              H                        ⁢            n                              
Namely, in an n-th eigenmode, an amplitude of noises change with an inverse of an n-th singular value, and SNR changes in proportion to a square of a singular value. It is widely known that there are also a minimum mean square error (MMSE) method and a maximum likelihood detection (MLD) method, as the method of determining the reception antenna weight.
The reasons why the method described above is used are as follows. The first reason is to deal with a temporal change of the channel matrix. Even if the channel matrix has changed prior to data transmission, it is possible to set the reception antenna weight matching the current channel matrix so that reception characteristics can be prevented from being degraded. The second reason is that feedback information can be made less. In orthogonal frequency division multiplexing (OFDM) communications such as those used in wireless LAN, it is necessary to set the transmission antenna weight for each subcarrier so that the amount of feedback information is very large. Since the feedback information is properly thinned from this reason, there may exist a large difference between the transmission antenna weight and the weight obtained through SVD of the channel matrix. However, since the reception antennal weight is set by considering the influence of the actually used transmission antenna weight, the characteristics can be prevented from being degraded.
FIG. 6 shows a probability distribution of a transmission (path) gain calculated from a square of a singular value. It is assumed that four transmission antennas and four reception antennas are used and each element of the channel matrix is an independent probability variable (Rayleigh fading) in conformity with the Rayleigh distribution. A transmission loss from each transmission antenna to each reception antenna is set to 0 dB. For reference, single-input single-output (SISO) with one transmission antenna and one reception antenna is also shown. SISO has an average transmission gain of 0 dB as the assumed condition shows. In contrast, in MIMO the fourth eigenmode has an average transmission loss of 8 dB, whereas the first eigenmode has an average transmission gain of about 10 dB. Therefore, for example, in a communication environment capable of obtaining a communication SNR of 30 dB, the first eigenmode can achieve an effective SNR of 40 dB. Therefore, in order to effectively utilize the first eigenmode, it is important to adopt modulation of a large number of levels and transmit a large amount of information.
However, for modulation of a large number of levels, a radio frequency (RF) circuit is required to have a high precision. Factors degrading a precision of an RF circuit include IQ mismatch, power amplifier non-linearity and the like. At a precision of a circuit currently used in wireless LAN, a limit of modulation is up to 64 QAM modulation, and 256 QAM modulation or higher is very difficult. Therefore, IEEE802.11a and 11g adopt only four schema BPSK, WPSK, 16 QAM and 64 QAM and cannot adopt 256 QAM. Even IEEE802.11n incorporating MIMO has a policy of not using 256 QAM. Therefore, even if a high SNR can be achieved, the number of levels of modulation cannot be made large, not leading to expansion of a communication capacity.
A method of solving this problem is proposed in JP-A-2005-323217 and “Studies on transmission method considering non-linear strain in MIMO-OFDM” by Yasuhiro TANABE, Hiroki SHOUGI, Hirofumi TSURUMI, 2005, The Institute of Electronics, Information and Communication Engineers, Composite Meetings, B-5-79. With this method, in a MIMO-OFDM wireless communication system for eigenmode transmission, the same modulation level is used for all subcarriers and all eigenmodes, and outputs of an error-correcting coder are sequentially assigned to different subcarriers. During this assignment, the eigenmode is changed for each subcarrier. With this method, the modulation number of levels of modulation becomes too large relative to the singular value in the eigenmode having a small singular value so that errors occur frequently. Conversely in the eigenmode having a large singular value, the number of levels of modulation is small relative to the singular value so that errors are hard to occur. Therefore, the error-correcting process corrects errors occurred in the eigenmode having a small singular value, and communications with a large communication capacity is possible.